JPH04119467A - Natural language translating method - Google Patents

Abstract

PURPOSE:To exactly process even a sentence with an indistinct tense expression by reading the sentence (input sentence) of a first natural language and converting it to a second natural language while changing it based on the corresponding relation of the tense expression between the input sentence and a translated result in a sentence translated before the input sentence. CONSTITUTION:For the sentence of the first natural language inputted from a source language input part 1, word division and syntax analysis is executed at a language analysis conversion part 2, and a tense decision part 3 decides the tense to be translated of the attention sentence while referring to the tense decision result of the preceding sentence stored in a tense decision result storage buffer 8 and stores the result in the tense decision result storage buffer 8. A language generation part 4 generates the sentence of the second natural language by using a language generation dictionary 7 and outputs it from an objective language sentence output part 5 afterwards. Thus, even the sentence with the indistinct tense expression can be exactly processed.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention provides a natural language translation method, in which a sentence in a first natural language (original language) read from a human-powered device is translated into a second natural language (original language). The present invention relates to a processing method in a device that automatically translates and outputs sentences in a target language, and particularly relates to a natural language translation method that can accurately process sentences with unclear tense expressions.

[Background Art] In recent years, with the development of computer technology, many automatic natural language translation systems have been proposed.

In conventional natural language automatic translation systems, when translating from a source language to a target language, tense attributes are usually analyzed for each sentence and translated based on the results.

Regarding this, for example, in the case of the original language or English, there is a method that analyzes the temporal relationship between events included in multiple sentences that make up a story, or by F., S.
“The Interpretation of T.
emporalRelations in Nara
tjve” A A A I '1988.

[Problem to be solved by the invention]

” In the two series of proposals, the main purpose is to interpret the story.
Because the tense attribute of a sentence is analyzed regardless of the tense attribute of the previous sentence, there were cases where the correct tense flow could not be translated.

For example, the source language is Japanese, the target language is English, and the previous sentence is that Company A will establish a new company in August. ...(1), and the sentence we are focusing on is: Company B will also participate in the new company. ...Consider the case (2).

In this case, the previous sentence (]) is extracted as a future nuance from the sentence element ``in August'' expressing time and the predicate ``to be established'', so the translated sentence is CO+n p a n
y ha l1lill establish a
nebucompany+, nAugust.

It is translated into the future tense.

However, since the sentence of interest (2) does not have a sentence element that expresses time, it is usually written as Company B also 1. akes par
t, in t, the new company.

This is translated into the present tense, but this is correctly translated as Comp.
Any Buuill Also Takespar
t+,n the new company.

should be in the future tense to match the previous sentence.

The present invention has been made in view of the above circumstances, and its purpose is to solve the problem of "double series" in the conventional technology, and to make it possible to accurately process sentences with unclear tense expressions. The purpose is to provide a language translation method.

[Means for Solving the Problems] The above-mentioned object of the present invention is to read a first natural language sentence (input sentence), analyze the read sentence, and then analyze the input sentence based on the result of the analysis. In a natural language translation device having a means for converting into a second natural language, the correspondence relationship of tense expressions between the input sentence and the translation result is determined based on the input sentence and the translation result in a sentence translated before the input sentence. This is achieved by a natural language translation method characterized in that the conversion into the second natural language is performed while changing the tense expression based on the correspondence relationship between the tense expression and the second natural language.

[Operation 1] In the natural language translation method according to the present invention, when analyzing the tense attribute of a sentence of interest, by referring to the tense attribute of a sentence translated before the sentence of interest, Even for sentences whose tense cannot be determined correctly just by looking at them, it is now possible to correctly translate the tense to match the flow of the sentence.

〔Example〕

Embodiments of the present invention will be described in detail below with reference to the drawings.

FIG. 2 is a diagram showing the basic block configuration of the entire automatic translation system that is an embodiment of the present invention.

In the figure, 1 is a source language input unit for manually inputting sentences in the source language, and 2 is a language analysis conversion dictionary 6 for converting input sentences in the source language.
a language analysis conversion unit that converts into other appropriate expressions using
4 refers to the contents of a tense determination result accumulation buffer 8, which will be described in detail later, and determines the tense to be translated for the sentence of interest; 4, based on the tense determined by the tense determination unit 3; A language generation section generates sentences in the target language using the language generation dictionary 7, and 5 indicates a target language output section that outputs the generated sentences in the target language.

Note that the tense determination result accumulation buffer 8 is as follows:
The identification numbers of translated sentences and their corresponding tenses are sequentially stored in a table.

An outline of the operation of the automatic translation system shown in this embodiment is as follows. In the following explanation, the source language is Japanese and the target language is English.

The Japanese sentence input from the source language input unit 1 is word-divided and syntax-analyzed by the language analysis conversion unit 2, and based on this result,
While referring to the tense determination result of the previous sentence stored in the tense determination unit 3 or the tense determination result accumulation buffer 8,
The tense to be translated for the sentence of interest is determined, and the result is stored in the tense determination result accumulation buffer 8.

From this result, after the language generation section 4 generates a sentence in the target language using the language generation dictionary 7, this is outputted from the target language sentence output section 5.

Hereinafter, the predicate of "ru form" - "established", the ending of the word "
Forms that express the past with the addition of ``Yu'' are called ``Yu-gata'' predicates, while forms that do not include ``Ri'', such as ``Establish'', are called ``Ru-form'' predicates. The specific operation of the above-mentioned tense determining section 3 will be explained by taking as an example a case where it is determined whether the tense is the future tense or the future tense.

The sentence to focus on is shown above. Eight companies will participate in the new company.

shall be. At this stage, just before this sentence, Company A will establish a new company in August.

The sentence has already been processed, the tense of the sentence is determined to be future tense, and the determination result is stored in the above-mentioned tense determination result storage buffer 8.
wcompany in August.

It is assumed that this is translated into the future tense.

The language analysis conversion unit 2 selects the sentence of interest. Eight companies will participate in the new company.

It is recognized that the predicate ``participate'' is in the ``ru form'' mentioned above, so for now, let's change the tense to the present tense and use the basic English translation as ``Company B also takes part.''
i,n the new company.

is obtained. Here, as shown in the operation flowchart of FIG. 1, the tense determination section 3 first examines the predicate (step 11), and since the r-participating J is in the "ru" form, the tense determination result accumulation buffer 8 is Examine the contents (step 12)
). As a result, the immediately preceding sentence is determined to be future tense (step 13), so the process proceeds to step 14, where the tense of the sentence of interest is changed to future tense. Based on this result, the language generation unit 4
so takes partol the new c
company.

Generates a translated sentence. The tense determination unit 3 stores this determination result in the tense determination result accumulation buffer 8, and ends the process.

Note that if the predicate is not in the ``le form J'', the tense change as described above is not performed (step 16).

] According to the twenty-first embodiment, even for sentences whose tense cannot be determined correctly just by looking at one sentence, the tense can be translated correctly to match the flow of the sentence.

FIG. 3 shows another embodiment. In this example, the previous sentence is `` Company 0 held a lecture yesterday. ...(3), and the sentence to focus on is: There were approximately 50 participants. Suppose that (4) is a formal expression.

At this stage, the previous sentence (3) has already been processed,
It is assumed that the tense of the sentence is determined to be past tense, and the determination result is stored in the double tense determination result accumulation buffer 8.

The language analysis conversion unit 2 selects the sentence of interest. There were approximately 50 participants.

Since the predicate ``about 50 people'' is recognized as a denotative tense, a translated sentence is generated by setting the tense to the present tense for the time being. Here, as shown in the operation flowchart of FIG. 3, the tense determination unit 3 first examines the predicate (step 21), and
Since "about 50 people" is a nominal limit, the contents of the tense determination result accumulation buffer 8 are checked (step 22). As a result, the immediately preceding sentence is determined to be past tense, so in step 23, the tense of the sentence of interest is changed to past tense.

The language generation unit 4 generates an English translation of the past tense based on this result. The tense determination unit 3 stores this determination result in the tense determination result accumulation buffer 8, and ends the process.

Note that if the predicate is not a denotative tense, no change in tense such as "double" is performed (step 25).

According to the above embodiment, even for sentences whose tense cannot be determined correctly just by looking at one sentence, the tense can be translated correctly to match the flow of the sentence.

It goes without saying that each of the above-mentioned Examples shows an example of the present invention, and that the present invention is not limited thereto. For example, in the embodiment described in section 1, the tense determination result accumulation buffer 8 was described as storing the identification numbers of translated sentences and their corresponding tenses in a sequential manner in a table. Result accumulation buffer 8
may be configured to remember at least the tense of the most recently translated sentence.

[Effects of the Invention] As described above in detail, according to the present invention, a first natural language sentence (input sentence) is read, the read sentence is analyzed, and the above-mentioned In a natural language translation device that has means for converting an input sentence into a second natural language, the correspondence relationship between tense expressions between the input sentence and the translation result is determined by determining the correspondence relationship between tense expressions in a sentence translated before the input sentence. A natural language translation method that can accurately process even sentences with unclear tense expressions, since the sentences are converted into the second natural language while being changed based on the correspondence of tense expressions between sentences and translation results. This has the remarkable effect of realizing the following.

[Brief explanation of drawings]

FIG. 1 is an operation flowchart showing the processing contents of the tense determination unit of a natural language translation device according to an embodiment of the present invention, FIG. 2 is a diagram not showing the basic block configuration of the natural language translation device according to the embodiment, The figure is an operation flowchart I showing another embodiment of the tense determination section. 1: Source language input section, 2 Language analysis conversion section, 3: Tense judgment section, 4: Language generation section, 5 Target language section output section, 6: Language analysis conversion dictionary, 7: Language generation dictionary, 8 Tense judgment result accumulation buffer. Diagram Original Language Sentence Diagram/Etese]~21

Claims (3)

[Claims] (1) A natural language having means for reading a sentence (input sentence) in a first natural language, analyzing the read sentence, and converting the input sentence into a second natural language based on the result of the analysis. In the translation device, while changing the correspondence relationship of tense expressions between the input sentence and the translation result based on the correspondence relationship of tense expressions between the input sentence and the translation result in a sentence translated before the input sentence, A natural language translation method characterized by converting into a second natural language. (2) When the first natural language is Japanese, and regarding the correspondence of tense expressions between the input sentence and the translation result, if the predicate is in the "le form", whether or not to make it the future tense. 2. The natural language translation method according to claim 1, further comprising determining: . (3) The first natural language is Japanese, and regarding the correspondence of tense expressions between the input sentence and the translation result, the tense expression is determined when the predicate is a nominal stop. The natural language translation method according to claim 1.